Every CSR row in residential trades loses calls. The good shops know which ones, why, and what to do about it by Tuesday morning. The shops running 65% booking floors lose 35% of calls and learn nothing because the audit lives in a manager's head, a binder of complaint emails, or the random one-call-per-week the owner happens to listen to. The "why did we lose this call?" audit closes the loop: AI transcribes and tags every lost call this week, surfaces the five most common kill reasons, builds the CSR coaching card in 20 minutes, and the floor rebuilds one objection line in the script by Tuesday morning. The named workflow: pull yesterday's call audit at 4 p.m., huddle the CSRs at 4:30, fix one objection in the script by Tuesday morning. This lesson is the build instruction — the audit query, the tagging taxonomy, the coaching-card template, the Monday tuning cycle, and the cadence that compounds booking-% lift weekly instead of waiting for the quarterly review that never happens.

Why the Call Audit Is the Coaching Loop the Shop Has Never Actually Run

Every shop owner says they coach the CSR row from call recordings. Almost no shop owner actually does. The reason is mechanical: pre-AI, listening to call recordings was a 4-6 minute investment per call to extract one coaching insight, multiplied by 30-40 lost calls per week, multiplied by 47 weeks of coverage — roughly 90-140 hours per year of dedicated listening time the owner or service manager does not have. The result in practice is "coaching" that consists of the floor lead's anecdotal sense of which CSR struggles, a random 3-4 calls reviewed per month, and the assumption that the rebuttal library handles the rest. Power Selling Pros, Service MVP / Joe Crisara, and the CertainPath coaching org built their consulting businesses on filling this gap with in-person scripts and binder content because the in-shop audit cycle was operationally impossible.

AI changes the math. Whisper-class transcription costs roughly $0.006 per minute of call audio in 2026; a 7-truck shop's 30-40 weekly lost calls of 4-6 minutes each is $5-$10 of transcription cost per week. Frontier LLMs (Claude, GPT-4-class) tag those transcripts against a custom rubric in seconds; CallRail Conversation Intelligence, ServiceTitan call summaries, Avoca's post-call notes, and Housecall Pro AI Team ship this capability natively. What previously cost 90-140 hours of manager listening time now costs $300-$500 per year in cloud spend and 20 minutes per week of audit review time. The audit is no longer a gap — it is the highest-leverage operational discipline available to the CSR-floor lead.

The 20-minute audit closes the loop the Lesson 1 rebuttal library and the Lesson 2 call summary workflow opened. Lesson 1 built the library; Lesson 4 tunes it from observed deflected calls. Lesson 2 built the summary infrastructure; Lesson 4 leverages the summary's AI-tagged kill-reason field to populate the audit without manual classification. The chapter's four lessons compound: rebuttal library + call summaries + after-hours capture + audit-driven coaching = the booking-% lift from 65 to 85 reproduces itself month over month rather than stalling at v1.

The Tagging Taxonomy That Makes the Audit Actionable

The single highest-leverage configuration choice in the audit workflow is the tagging taxonomy — the set of kill-reason categories the AI uses to classify each lost call. Vendor-default taxonomies are too broad ("not booked," "price objection") to drive coaching; the shop's calibrated taxonomy maps to specific rebuttal-library entries and specific coaching cards. The taxonomy is 12-15 categories, mutually exclusive, and tuned from the shop's actual lost-call patterns over the first 60 days of audit cadence.

The Five Headline Kill Reasons That Cover 80% of Lost Calls

The five headline kill reasons across residential trades CSR rows in 2026 cover roughly 80% of lost calls. Price-shock: homeowner heard the dispatch fee or the rough cost estimate and disengaged before the rebuttal landed. Slot-mismatch: homeowner needed a window the dispatch board could not offer and the CSR did not pivot to a backup slot or a held-with-text-confirm. Service-area-out: homeowner was outside the shop's service area or in a marginal zip the CSR was not authorized to commit to. No-rebuttal-attempted: the CSR encountered an objection and did not pull from the rebuttal library — typically a CSR new to the floor or a CSR who did not have the library loaded. Rebuttal-deflected: the CSR attempted a rebuttal but the homeowner pushed back and the CSR did not have second-pass language available. Each kill reason maps to a specific coaching response, a specific library update, and a specific 4 p.m. review surface.

The Secondary Kill Reasons

The remaining 20% breaks across 7-10 secondary kill reasons. Language-barrier: homeowner's primary language is not English and the CSR's Spanish or other language fluency was insufficient. Complex-routing: homeowner had a multi-property or multi-system situation the CSR could not triage in confidence. Warranty-confusion: prior service or warranty status the CSR could not resolve in the call window. Spouse-decision: homeowner deferred booking to a partner who was not available, and the CSR did not pivot to a held-slot-with-text-confirm. Hostile-caller: caller was angry from a prior interaction or a competitor's failure and the CSR did not de-escalate before booking. Misroute-from-AI: the AI receptionist warm-transferred a call to the human CSR that the AI should have handled, and the handoff fumbled. Tech-preference: homeowner asked for a specific tech who was not available and the CSR did not pivot. Time-of-day: caller wanted same-day service the shop could not honor and the CSR did not pivot to next-morning held-slot.

The taxonomy is mutually exclusive — every lost call gets exactly one tag — and weekly tuning catches drift. If a new kill-reason category emerges with 4+ instances in a week, the floor lead adds it and re-runs the prior month's audit against the expanded taxonomy. The taxonomy compounds in precision over time as the shop's lost-call patterns surface and the rebuttal library tunes against them.

The Named Workflow: The 4:30 P.M. Huddle and the Tuesday Morning Fix

The audit operates as a daily cadence with a weekly tuning beat. The named workflow has four anchor points: 4:00 p.m. audit pull, 4:30 p.m. floor huddle, weekly Monday lead-review, and Tuesday morning script update. Each anchor takes a specific role and produces a specific artifact.

4:00 p.m. — Audit Pull

The CSR-floor lead pulls the day's lost-call audit from the AI summary engine (CallRail Conversation Intelligence, ServiceTitan call summaries, Avoca's post-call notes, or Housecall Pro AI Team). The pull is a dashboard query that returns the day's calls tagged with kill reason, CSR identity, time of call, duration, and a link to the recording. Typical volume: 20-35 lost calls per day at a 7-truck shop's CSR row. The lead reads the kill-reason distribution — which categories dominated the day — and identifies the three most-deflected calls for 60-second listen-back. Total time: 10 minutes.

4:30 p.m. — Huddle

The CSR floor convenes for a 10-minute end-of-shift huddle. The lead walks the floor through: (1) today's kill-reason distribution (one slide or whiteboard column), (2) the three most-deflected calls (15 seconds of audio per call), (3) the coaching insight from each (what the CSR could have said instead), and (4) the one rebuttal-library line that needs tightening before tomorrow morning. The CSRs absorb the coaching in real-time with the audio fresh. Total time: 10 minutes. The huddle is end-of-shift, not start-of-day, because the day's calls are fresh and the lead has had a day to identify patterns.

Monday — Lead Review

The CSR-floor lead aggregates the prior week's kill-reason distribution into a weekly cut. The cut shows: total lost calls by category, week-over-week trend, per-CSR distribution (without naming for the huddle — the lead absorbs per-CSR signal privately), and the rebuttal-library lines that produced the most deflections. The lead drafts the one specific library update or coaching card the floor needs by Tuesday morning. Total time: 20 minutes. The Monday lead review is the depth check that the daily 4 p.m. audits don't have time for.

Tuesday Morning — Script Update

The Tuesday morning huddle (the Lesson 1 cadence) rolls out the one library update or coaching card identified in the Monday review. The laminated card, the digital reference, and the AI voice agent's system prompt all update simultaneously. Every CSR practices the new language out loud once. The change is live by 8 a.m. Tuesday. By Friday, the floor has 4 days of data on whether the new language is landing or whether it needs further tuning the following Monday. The cadence compounds 1-2 booking-points per month as documented in Lesson 1.

The four anchors are non-negotiable as a sequence. Skipping the 4 p.m. audit drops the floor blind on the day's pattern. Skipping the 4:30 huddle leaves the coaching abstract instead of audio-grounded. Skipping the Monday review collapses tactical daily insight into anecdote without weekly synthesis. Skipping the Tuesday update means the floor learns the same lesson repeatedly without operationalizing the fix. Most shops that have tried call audits and stalled have one of these four anchors missing.

The Coaching Card Template

The audit produces one artifact each Tuesday morning: the coaching card. The card is a single-page document that captures the prior week's most actionable insight and the script update for the coming week. The template has six fields, designed to be skimmable in 90 seconds and actionable in the next call.

Card Field One: The Pattern

One sentence describing the kill-reason pattern surfaced in the prior week. Example: "Price-shopper rebuttal deflected 14 times this week when the homeowner pushed back with 'but my neighbor said it should only be $89 to come out.'" The pattern names the kill-reason, the volume, and the specific homeowner language that defeated the rebuttal. Without the specificity, the card devolves into generic advice; with it, the CSR can pattern-match the next call instantly.

Card Field Two: The Prior Rebuttal

The current rebuttal-library language that failed against the pattern. Example: "'Here's the honest answer: the only way to give you a real number is to put a licensed tech on it — the $79 dispatch covers a full diagnostic and gets credited against any repair you approve.'" Quoting the prior rebuttal makes the failure mode concrete — the CSR sees exactly what they had been saying and understands what needs to change.

Card Field Three: The New Rebuttal

The Tuesday morning replacement language, drafted from the most-deflected calls. Example: "'Totally — and your neighbor is probably remembering a shop that's calling it a $89 service charge with no diagnostic included; ours covers the actual diagnostic and credits the full $79 toward any repair you approve. The cheap dispatch is usually the shop that surprises you with the labor charge once they're at the panel.'" The new rebuttal directly addresses the deflection language the homeowner used. Drafted by the floor lead in the Monday review; rolled out Tuesday morning.

Card Field Four: The Practice Script

Two example dialogues showing the new rebuttal in context: one where the homeowner accepts it and books, one where the homeowner pushes back further and the CSR uses the second-pass language. The practice script lets the CSR rehearse the language out loud in the Tuesday huddle and arrive at the call ready to use it.

Card Field Five: The System-Prompt Update

The exact text update for the AI voice agent's system prompt. The CSR-floor lead pastes this into Avoca, Jobber AI Receptionist, HCP AI Agents, or ServiceTitan Voice on Tuesday morning before 8 a.m. so the AI receptionist handles the same pattern with the same shop voice. Without this field, the human floor learns the new language but the AI keeps producing the old failed language on after-hours and overflow calls.

Card Field Six: The Success Metric

The specific number to watch over the coming week. Example: "Price-shopper-with-neighbor-comparison rebuttal-deflected count this week. Target: drop from 14 to under 4 by Friday." Tying the card to a measurable outcome gives the next Monday review a clean check on whether the update landed or needs further tuning.

The Per-CSR Distribution and the Coaching Conversation

The audit's per-CSR distribution is the highest-leverage and most-handled artifact in the workflow. The lead sees: which CSR's kill-reason distribution diverges from the floor average, which CSR has the highest rebuttal-deflected count, which CSR is missing specific objection types entirely. Per-CSR data drives individual coaching; floor-average data drives library tuning. Both matter; mishandling the per-CSR data destroys floor trust.

The discipline that protects floor trust: per-CSR data never surfaces in the 4:30 huddle. The huddle reads floor-average distribution and the three most-deflected calls anonymized. Per-CSR coaching happens one-on-one in scheduled 15-minute weekly conversations, owned by the floor lead, anchored to specific calls the CSR will hear. The conversation focuses on the kill-reason categories where the CSR's distribution diverges from floor average, walks through 2-3 specific calls, and identifies the language change or rehearsal pattern that fixes the divergence. Coaching is collaborative — the CSR offers the language they'd try next; the lead refines it.

Per-CSR coaching cadence: weekly 15-minute one-on-ones, owned by the floor lead, scheduled in advance, never the daily 4:30 huddle. The cadence is the discipline that makes the audit a coaching tool rather than a performance-tracking surveillance tool. Floors that surface per-CSR data in the huddle convert the audit into a public-shaming mechanism within 30 days and lose their best CSRs; floors that hold per-CSR data in the one-on-one cadence build trust and compound booking-% lift quarter over quarter.

The Metrics the Audit Moves and the 90-Day Rollout

The audit workflow moves four primary numbers and two diagnostic numbers. All are visible on the dashboard the floor lead reads at 7:00 a.m. every weekday morning.

Booking %. The end metric. Industry baseline 65%; target 80-85%. The audit lifts booking % indirectly by tuning the rebuttal library faster than the Lesson 1 cadence alone would produce. Shops running the full audit cadence land at 80-85% in 60-75 days versus 75-90 days for shops running only the Lesson 1 cadence. The audit is the velocity multiplier.

Rebuttal-deflected count. The middle metric. The number of calls per day where the CSR attempted a rebuttal and the homeowner deflected. Drops 50-70% over the first 60 days of audit cadence as the rebuttal library tunes faster. Healthy band: 3-7 per day at a 7-truck shop's CSR row by day 90.

Kill-reason distribution stability. The diagnostic metric for the audit's own health. Once stabilized, the kill-reason distribution should be steady week-over-week — price-shock 25-30%, slot-mismatch 15-18%, service-area-out 8-12%, no-rebuttal-attempted under 5%, rebuttal-deflected 20-25%, balance across secondary categories. Significant week-over-week drift signals either a tuning win (a category dropped) or a new pattern emerging (a category rose).

Tuesday morning script-update count. The cadence metric. The number of script updates the audit produced over the prior month. Healthy band: 3-5 per month. Zero means the audit isn't producing actionable insights or the lead is skipping the Monday review. Above 8 means the floor is over-rotating on tactical changes and not letting language settle.

Two diagnostic metrics: per-CSR rebuttal-deflected variance (the spread between the CSR with the highest deflected count and the CSR with the lowest — too high indicates coaching gap; too low indicates floor uniformity) and one-on-one cadence completion (every CSR receiving their weekly 15-minute one-on-one within 7 days of scheduling).

The 90-day rollout: Week 0, configure the AI summary tagging taxonomy with the five headline kill reasons and the eight secondary; train the floor lead on the 4 p.m. and Monday review cadences. Week 1, deploy. The audit runs daily; the 4:30 huddle starts. The first Tuesday morning script update lands. Weeks 2-4, kill-reason distribution stabilizes as the taxonomy tunes; rebuttal-deflected count begins to drop. Weeks 5-8, per-CSR coaching one-on-ones land in cadence; the floor lead's confidence in the data builds. Weeks 9-12, the workflow stabilizes; booking-% climbs to the 80-85% range as the audit-driven library tuning compounds. By day 90, the audit is the floor's operating cadence — the lead and the CSRs run it without supervision, and the booking-% lift becomes a structural feature of the shop rather than a quarterly initiative.

Key Takeaways

• The 20-minute weekly audit replaces 90-140 hours/year of pre-AI manager listening time at $300-$500/year in cloud spend. What used to be operationally impossible is now the highest-leverage operational discipline available to the CSR-floor lead.
• Five headline kill reasons cover 80% of lost calls: price-shock, slot-mismatch, service-area-out, no-rebuttal-attempted, rebuttal-deflected. Eight secondary categories cover the remaining 20%. Mutually exclusive; weekly tuning catches drift.
• The named workflow has four anchors: 4:00 p.m. audit pull (10 min), 4:30 p.m. huddle (10 min), Monday lead review (20 min), Tuesday morning script update (rolls into the Lesson 1 weekly cadence). Skipping any anchor breaks the loop.
• The coaching card has six fields: the pattern, the prior rebuttal, the new rebuttal, the practice script, the system-prompt update for the AI voice agent, the success metric for the coming week. One page, 90-second skim, actionable in the next call.
• Per-CSR data never surfaces in the 4:30 huddle. The huddle reads floor-average distribution and anonymized most-deflected calls. Per-CSR coaching happens in scheduled weekly 15-minute one-on-ones. Floors that mishandle this destroy trust and lose their best CSRs in 30 days.
• The audit lifts booking-% velocity — not the ceiling. Shops with the full audit cadence reach 80-85% in 60-75 days; shops with only the Lesson 1 cadence reach the same ceiling in 75-90 days. The audit is the multiplier on top of the rebuttal library, not a substitute.
• Four primary metrics: booking % (65 → 85), rebuttal-deflected count per day (drops 50-70% in 60 days), kill-reason distribution stability (week-over-week steadiness once tuned), Tuesday script-update count (3-5/month healthy band).
• The audit closes the loop the chapter opens. Lesson 1 built the rebuttal library; Lesson 2 built the summary infrastructure; Lesson 3 built the after-hours capture; Lesson 4 tunes all three from observed lost-call data. The four lessons compound into structural booking-% lift that reproduces itself month over month.
• The coaching cadence is what protects the AI rollout from CSR-floor revolt. Audit data used as coaching builds trust; audit data used as surveillance destroys it. The discipline of one-on-one coaching versus public huddle reading is the difference between a 24-month rollout that compounds and a 90-day rollout that reverses.